Here we process reports by identifying specific character names in each sentence to account for character-specific sentiment analysis. We’ll examine over general storylines (across all chapters) and chapter-by-chapter karmic balance.
As for the overall team karma, using afinn lexicon, score each report over time using a cumulative sum. The reports all trend downward using raw scores, indicating an overall negative tone. To help account for this, there is a pos_adj parameter that calculates the ratio of abs(sum(neg) / sum(pos)). Multiplying the positive scores by (some fraction of) this parameter helps boost those scores in relation to negative scores.
For each character, chart out the cumulative sum of ‘karmic balance’ - using word count as a metric of time. This is averaged over all adventures to estimate the general character ethos - how they approach a mission.
A bar chart showing the karmic balance of each character for each chapter - resulting in a general character arc over the campaign.